Posted on 2008-06-02 12:38:42-07 by aalves
Compound Words?
Hi Everybody,

I've looked the new release (0.10) of SenseRelate::AllWords and realized that we don't presently have the option:

wsd.pl ... [--compounds FILE] ...

At this moment, I'm using the previous release (0.9) and, curiously, I'm facing problems with compounds words using Jcn (Jiang & Conrath) measure: In some experiments, this measure is unable to associate the meaning of compound words(neither recognize them as noun, at all) giving to wsd.pl the file containing all the compound words that is already known. Is there any problem with the use of this measure on compound words?


Examples:

Disambiguated Sentence using difference measures but with the same wsd.pl configuration:

scheme = normal, format = tagged, compounds = compounds.txt (this contains the words miami_beach and fast_food), stoplist = stoplist.txt

Sentence1: Every year I usually go to Miami Beach on vacation

Tagged Sentence: Every/DT year/NN I/PRP usually/RB go/VBP to/TO Miami/NNP Beach/NNP on/IN vacation/NN


jcn measure: Every year#n I usually#r go#v to miami_beach on vacation#n

lch measure: Every year#n I usually#r go#v to miami_beach#n#1 on vacation#n#1

lesk measure: Every year#n#4 I usually#r#1 go#v#1 to miami_beach#n#1 on vacation#n#1

res measure: Every year#n I usually#r go#v to miami_beach on vacation#n


Sentence2: At lunch I prefer slow food than fast food

Tagged Sentence: At/IN lunch/NN I/PRP prefer/VBP slow/JJ food/NN than/IN fast/JJ food/NN


jcn measure: At lunch#n I prefer#v slow#a food#n than fast_food food#n#1

lch measure: At lunch#n I prefer#v slow#a food#n#1 than fast_food#n#1 food#n#1

lesk measure: At lunch#n#1 I prefer#v#2 slow#a#1 food#n#1 than fast_food#n#1 food#n#2

res measure: At lunch#n I prefer#v slow#a food#n#1 than fast_food#n#1 food#n#3

Could anyone suggest me in what I'm wrong?
Thanks,
Ana Alves
ana@dei.uc.pt
Direct Responses: 8000 | Write a response
Perl Weekly newsletter
A free weekly newsletter for people who are busy to read all the blogs. click here to check it out.